Figure 3

De novo rediscovery of motifs and differential analysis. (a) Scheme on rediscovering the Dux motif by removal from the background motif database. Separate DENIS runs on a Dux positive and on a Dux negative condition. (b) Binding site enrichment (y axis) of motifs calculated by DENIS (x axis) as described in (a) within Dux ChIP-seq peaks. The top motifs enriched for Dux positive (green) and Dux negative (red) conditions are shown. (c) Scheme on how to discover motifs unique to the Dux positive condition without the use of any motif database. DENIS uses the differential FP module to predict motifs exclusively present in the Dux positive condition. (d) Differential motifs as generated in (c), sorted by distance (x axis) to their closest match in the motif database. Motifs below a distance of 0.4 are considered as rediscovered database motifs, motifs above are considered novel. (e, f) Examples of rediscovered motifs and their corresponding JASPAR database matches. (g) Combined differential motif activity plot of JASPAR and DENIS (novel & rediscovered) motifs. For each motif the fold change of activity between conditions (x axis) and corresponding pvalue is depicted. (h) Excerpt of motif similarity clustering as generated from TOBIAS based on binding site overlap. Motif 10 represents the rediscovered DENIS version of the Dux motif (see also (e)). (i, j) Exemplary TF-class assignment for one rediscovered (i) and one novel (j) motif. The number of JASPAR motifs with high binding sequence overlap to the respective de novo motif is depicted as N, and the total number of motifs within the JASPAR database is given as DN. The plots show enrichment (x axis) for assigned TF-classes (y axis, one row per class). The enrichment is computed by counting the number of motifs in N assigned to a TF-class (depicted as n) and comparing it to the count of motifs in DN assigned to the same class (depicted as dn). Parts (a) and (c) are created with BioRender.com.